Corporate finance and financial-modeling workbench for spreadsheets, analysis, advice, and model review. Covers EBITDA, ratios, common-size/trend analysis, 3-statement models, cash-flow statements, 13-week forecasts, working capital/CCC, DCF/WACC/valuation, and budgeting. Produces XLSX / Claude-for-Excel workbooks with assumption sheets, validation checks, color-coded inputs, auditable formula traces, and Python recalculation checks. Triggers on "build a financial model", "make a budget", "3-statement model", "DCF", "WACC", "valuation", "cash flow forecast", "ratio analysis", "EBITDA", "common-size", "P&L template", or model review. Every stated number needs source, formula, substitution, units, result, Python verification, and mismatch status. Skip for bookkeeping entries, tax filing, or personal-finance budgeting.
COBESY (Cognitive Behavioral Systemic) helps an agent turn static knowledge (docs, repos, architecture, decks) into work people actually adopt. It checks the culture first (Schein assumptions, Edmondson safety, Scientist Mode), then compresses the message (Minto, Dirksen, Knowles), then plans how adoption moves through the network (Centola, Jackson, Berger REDUCE, cascades). The same cognitive engine also structures written and projected artifacts: blog posts (technical and financial), a thesis, journal article, or slides, with a skeleton grid, headlines, bullets, emphasis, writing advice, style guidance, and keywords before prose (Belcher, Cron, Kolin, Graff & Birkenstein, Klinkenborg, Williams & Bizup). Use it for rollouts, buy-in, change management, briefings, resistance, onboarding, or to structure a piece of writing where the knowledge is there but the structure keeps getting lost. Skip it for factual lookup or single-fact Q&A.
Negotiation and auction playbook for bargaining, procurement, and incentive design. Covers reservation price, BATNA, ZOPA, surplus, aspiration ladder, degressive concessions, counter-offers, nibble defense, higher-authority commitments, virtual price lists, holdup protection, fairness checks, Vickrey mechanisms, negotiauctions, profit-share sizing, and draw.io maps of players, incentives, moves, and payoffs. Triggers on requests to prepare for or run a negotiation, choose an opening offer or concession pattern, defend a walk-away price, improve a BATNA, structure a tender, design a bonus, split something fairly, or visualize a negotiation as a game. Also fires on "help me negotiate", "BATNA", "ZOPA", "reservation price", "degressive bidding", "negotiauction", "game tree", "payoff matrix", and on salary, vendor/procurement, M&A earnout, and security-compliance negotiations. Skip for general game-theory homework unrelated to bargaining, auctions, or contracting.
General draw.io diagram creation via MCP tools. Covers the three tool types (Mermaid, XML, CSV), URL presentation rules for the Hermes WebUI, Mermaid syntax reference shortcuts, domain-specific palettes (executive/boardroom Gartner-style), and when to prefer each tool type. Use for any draw.io task that doesn't fit a domain-specific skill (intana-viz, architecture-diagram, etc.). Load this skill when the user asks for a draw.io diagram, a Mermaid diagram for draw.io, or any visualization using the draw.io MCP tools.
Create intelligence analysis visualizations using draw.io diagrams. Covers 80+ structured analytic techniques including: ACH, Argument Mapping, String of Pearls, Critical Path, CoA Analysis, Pre-Mortem, Red Teaming, Four Ways of Seeing, Devil's Advocacy, Prioritisation Maps, If/Then matrices, Bias catalogs, Opportunity matrices, McKinsey 7S, Cultural Web, Leavitt's Diamond, Fishbone, Futures Wheel, Six Thinking Hats, Key Assumptions Check, Blindspot Analysis, Framestorming, War Gaming, and more. Use this skill whenever the user asks for any intelligence analysis chart, diagram, matrix, framework visualization, structured analytic technique visualization, or analytical model rendering — even if they don't name a specific technique. Also trigger for risk assessments, decision matrices, stakeholder analysis, assumption mapping, scenario planning, wargaming, root cause analysis, or any structured thinking framework.
Unified SOC / DFIR workflow rooted in Investigation Theory (Diagnostic Inquiry): six-stage loop, question taxonomy (preceding / context / succeeding / proximate / capability-matching / utility), GAPSS data manipulation, three-tier escalation, modified CJCSM 6510 dispositions, Once Upon a Time compromise report, Security M&M peer review, and five standard draw.io diagrams. Triggers on alert triage, log analysis (proxy / mail / Windows / Sysmon / EDR / NetFlow), phishing, lateral movement, C2 / beaconing, data exfiltration, malware execution, insider threats, and on requests to write case notes, playbooks, attack timelines, dispositions, compromise reports, or run an M&M peer review. Also fires on prompts like "help me investigate this alert", "how do I analyze these logs", "build me a playbook for X", "what questions should I ask", "document this incident", or mentions of SOC workflows, triage queues, or incident severity. Does not fire on detection engineering or offensive / red-team work.
Stress-test a concrete plan, launch, hire, pricing change, strategy, or commitment by assuming it has already failed and reasoning backward. Use when the user has a real decision on the table and asks to 'premortem' it, 'stress-test' it, 'find blind spots', 'poke holes', or otherwise wants honest failure analysis instead of validation. Skip for vague ideas (help them plan first), questions with one right answer, or decisions that are already irreversible.
Turn a rough user request into a clear prompt using ReFLECT: Role, Format, Language, Example, Context, Task. Ask only for missing pieces that matter: always a missing Task, and missing Context or Example when the task depends on them. Use when the user says 'reflect this', 'apply ReFLECT', 'turn this into a good prompt', 'engineer this prompt', or gives a vague request they want cleaned up before it runs. Skip complete requests and do the work directly.